Shingled Magnetic Recording (SMR): What It Means for a NAS

Shingled magnetic recording (SMR) is a hard drive recording method that overlaps each new track on part of the previous one, like roof shingles, to fit more data per platter. The trade-off: changing data in place means rewriting a whole band of tracks, so sustained writes and RAID rebuilds slow badly. For NAS array members, buy CMR.

SMR is not a defect, and it is not slow at everything: it reads about as fast as a conventional drive, and it is how vendors build their biggest disks. The catch is that drive-managed SMR hides its rewrite cost until the day you replace a failed disk. Plenty of homelabbers learn the term the expensive way, halfway through a rebuild.

How does shingled magnetic recording work?

SMR writes tracks that partially overlap, so the drive needs no guard gap between them. Per Western Digital’s Zoned Storage documentation, the write head is wider than a single track, and after writing one track the head advances by only part of its width, covering a strip of the track it just wrote.

That buys areal density, and it creates one hard rule. Overlapping tracks are grouped into fixed-capacity bands called zones, with gaps between zones so the wide write head can’t clobber the next band. Inside a zone, data has to be written sequentially. Change one sector in the middle of a zone and, per the same documentation, the entire zone must be rewritten, because rewriting that track would trample the tracks shingled on top of it.

The roof analogy holds up. Laying shingles from the bottom row upward is easy; replacing one in the middle means lifting everything above it. CMR (conventional magnetic recording) leaves gaps between tracks, so any sector can be overwritten in place with no knock-on work. Where an SMR drive enforces the sequential rule is the whole story for a NAS owner.

What is the difference between drive-managed, host-aware and host-managed SMR?

The three SMR types differ in who enforces the sequential-write rule. Drive-managed SMR handles it in firmware and looks like a normal disk. Host-managed SMR requires the operating system to write each zone sequentially. Host-aware sits in between: the drive accepts random writes, but exposes its zones so software can optimize around them.

The zone models and command sets are documented on the Zoned Storage SMR page; the TrueNAS row comes from the TrueNAS Hardware Guide.

Drive-managed is the one that matters for a home NAS, because it is the one sold without a warning label. It presents a backwards-compatible interface, so your NAS OS formats it, adds it to the array, and never knows anything is different.

Why do SMR drives struggle in a NAS rebuild?

Drive-managed SMR drives hide the rewrite penalty behind a cache, and a rebuild is the workload that exhausts it. Incoming writes land in a CMR-style staging area and get moved to shingled zones during idle time. A resilver is a long stretch of sustained writes with no idle time, so the drive ends up rewriting zones while the array is degraded.

ServeTheHome’s write-up describes that cache mechanism, and Western Digital effectively confirmed it. In its April 2020 post, WD disclosed that its 2TB to 6TB WD Red drives used drive-managed SMR and argued that home NAS workloads are “intermittent, leaving sufficient idle time” for background data management. Its June 2020 update conceded that sustained random writes during ZFS resilvering cause “a lack of idle time” and “significantly lower performance.”

The numbers that made the rounds come from ServeTheHome’s May 2020 comparison. They used each 4TB drive as the replacement disk in a 4-drive RAIDZ pool filled to around 60 percent, while also copying files and reading data over the network. Per the results page, the CMR WD40EFRX, the Seagate IronWolf and the HGST drive all finished in under 17 hours. The SMR WD40EFAX took nearly 230 hours, roughly nine and a half days of running without full redundancy.

Here’s the nasty part. In the same article, the SMR drive read “fairly similarly” to its CMR sibling and beat it in PCMark8, likely helped by its 256MB cache versus 64MB. Most quick benchmarks you’d run on a new disk look fine. Only the file copy test, and CrystalDiskMark run straight after heavy writes, showed it falling behind.

iXsystems found something worse than slow. Its TrueNAS compatibility notice says the WD40EFAX on firmware 82.00A82 could enter a faulty state under heavy write loads, including resilvering, returning IDNF errors that ZFS treats as a drive failure. iXsystems stopped shipping those drives in its Mini systems. A replacement disk that faults mid-resilver while you’re already one drive down is how you end up doing a full restore. You do have the backup, right?

ZFS got the headlines, but any rebuild that writes to a replacement disk for hours is the same sustained-write, no-idle-time workload WD’s explanation describes.

Where is SMR actually fine?

SMR is fine where data is written once, mostly sequentially, and rarely rewritten. Seagate’s CMR/SMR list says SMR is “best suited for workloads that are sequential or easily sequentialized,” naming large media repositories, backup tiers, compliance archives and object storage. For a homelab, that means a standalone backup disk, not an array member.

The same Seagate page shows where SMR still lives: 3.5-inch BarraCuda 4TB and 8TB, 2.5-inch BarraCuda 2TB and 4TB, 2.5-inch FireCuda 1TB and 2TB, the 8TB Archive drive, and Exos at 32TB, 36TB and 44TB. The 2.5-inch entries matter if you’re buying small bare drives for a mini-PC or a portable backup enclosure. The Exos entries explain why SMR isn’t going away: Seagate says SMR “enables the highest capacity hard drives available,” while noting that deploying it takes control over the software stack, zone management and compatible filesystems.

That caveat points at zone-aware software, and Linux has more of it than most people expect. The Zoned Storage kernel overview dates zoned block device support and native f2fs support to kernel 4.10, zonefs to 5.6, Btrfs zoned mode to 5.12 and native XFS support to 6.15. The Btrfs zoned mode docs list real limits: only the single and DUP profiles are supported, and zoned mode won’t work with DM-SMR disks.

So a cheap secondhand host-managed drive can be a fun Btrfs or zonefs experiment on a spare box. It is not a ZFS pool member, and it’s more work than it’s worth unless you already wanted to learn zoned storage.

How can you tell if a drive is SMR?

Match the exact model number against the manufacturer’s CMR/SMR table. That is the only reliable check for drive-managed SMR, because capacity, RPM and family name don’t settle it: the WD40EFAX is SMR while the older WD40EFRX is CMR, and both are 4TB WD Reds. Linux can only flag host-aware and host-managed drives.

Three commands get you most of the way on a Linux box or a NAS with SSH enabled:

# 1. The model number: match this against the vendor's CMR/SMR table
sudo smartctl -i /dev/sda | grep 'Device Model'

# 2. The zone model: "none" means CMR *or* drive-managed SMR
cat /sys/block/sda/queue/zoned

# 3. A weak hint only: a spinning disk that advertises TRIM is often DM-SMR
sudo hdparm -I /dev/sda | grep -i trim

The second check is the one people over-trust: the zoned block device docs state that drive-managed SMR disks report none, just like a conventional drive. The TRIM hint comes from a contributor in OpenZFS issue 10182, who argued in April 2020 that a rotational device reporting TRIM is “definitely SMR - or at least using zones internally,” citing the WD EFAX drives. Treat a TRIM hit as a reason to look up the model; no TRIM proves nothing.

Behavior is the other clue. A long copy that starts at full speed, then drops sharply and stays slow, fits the cache-exhaustion pattern, but so do several network and controller problems. Rule those out with the NAS slow transfer diagnosis before blaming the platters. For per-model buying checks, see the CMR vs SMR NAS drive guide.

Should you put SMR drives in a home NAS?

No, not as array members. Buy CMR for anything in a mirror, RAIDZ, SHR or any other parity array, and treat SMR as acceptable only for a standalone backup disk you fill sequentially and mostly leave alone. A modest capacity or price advantage is not worth a rebuild measured in days.

On the CMR side, Western Digital’s own post points ZFS and heavier NAS users to WD Red Plus and WD Red Pro, and Seagate’s list puts its SMR models in BarraCuda, FireCuda, Archive and the largest Exos capacities rather than IronWolf. Still check the exact model and your NAS vendor’s compatibility list; the NAS buying guide covers that alongside bay count and controller choices. For CMR drive picks for a ZFS pool, see best hard drives for TrueNAS.

Already have one in the array? Don’t yank it in a panic; pulling a healthy disk creates the very rebuild you were worried about. Restore a test file to prove your backups, check pool health, then replace SMR disks one at a time, on your own schedule, with documented CMR models. Swap any SMR cold spare first. A spare has exactly one job, and it’s surviving a rebuild.

FAQ

Is shingled magnetic recording bad?

No, shingled magnetic recording is not bad in general; it is a capacity trade-off. SMR drives read normally and suit sequential, write-once data such as archives and backup tiers. They are a poor fit for NAS arrays, where rebuilds and frequent overwrites hit the zone-rewrite penalty that drive-managed models hide until their cache runs out.

Is SMR slower than CMR?

SMR is slower than CMR at sustained writes, not at everything. ServeTheHome’s 2020 comparison found the SMR WD40EFAX read about as fast as the CMR WD40EFRX and even edged it in PCMark8, but it was significantly worse in the file copy test and took nearly 230 hours to resilver, against under 17 hours for CMR drives.

Does ZFS support SMR drives?

Not well. The TrueNAS Hardware Guide says drive-managed and host-aware SMR drives function poorly with TrueNAS and are not recommended, while host-managed SMR drives do not work at all because ZFS cannot manage their zone semantics. For ZFS pools on TrueNAS, Proxmox or plain Linux, buy documented CMR drives instead.

Is an SMR drive OK for backups?

Yes, for a single-disk backup that is written mostly sequentially and then left alone. Seagate lists backup tiers and archives among the workloads SMR suits best. Avoid it as a backup target that takes heavy rewrites, and run a restore drill: a completed copy and a test restore prove more than the drive label does.

Related across the network

• Best Hard Drives for Unraid: CMR Picks by Budget (2026) — unraidhub.com
• Best Budget NAS Build for Home: What Actually Works in 2026 — truenasguide.com